Nvidia is developing Nemotron 4, an open AI model with at least 1 trillion parameters, aiming to drive GPU demand and expand its open AI ecosystem. The company is also releasing a lightweight open model for single-GPU operation, targeting startups and general companies.
TL;DR
- Nvidia's Nemotron 4 will have at least 1 trillion parameters, nearly twice the size of its predecessor.
- The model is part of Nvidia's strategy to expand GPU demand among startups and enterprises.
- Nvidia is also releasing a lightweight open model, Nemotron 3.5 Lightning, for free use and modification.
What happened
Nvidia is developing Nemotron 4, an open AI model with at least 1 trillion parameters, according to The Information. The model is expected to be completed as early as late this fall and will challenge the world's best open-source models. The largest model in the Nemotron 3 series, Nemotron 3 Ultra, has 550 billion parameters and scored 48 points on the Intelligence Index, trailing behind top closed and open models like Anthropic's Claude Opus 4.8 and Moonshot AI's Kimi K2.6.
Alongside Nemotron 4, Nvidia is releasing Nemotron 3.5 Lightning, an open model with 30 billion parameters (active 3 billion) that can be downloaded and modified for free. This lightweight model uses distillation technology to specialize in AI agents' repetitive tasks and is part of Nvidia's strategy to expand GPU demand. The company is also introducing the open-source routing tool Nimo Switchyard, which automatically allocates requests to the most suitable and cost-effective model, reducing AI token costs for companies.
Why it matters
Nvidia's focus on open models is driven by the belief that cheaper, open models will increase demand for its GPUs and data centers. CEO Jensen Huang has previously stated that free AI is beneficial for hardware and chips. The strategy also aims to diversify GPU consumers, as many large customers are developing their own AI chips. By targeting startups and general companies with low-cost open models, Nvidia can expand its customer base and drive GPU demand.
The development of Nemotron 4 is also a result of Nvidia's Nemotron Alliance, formed in March with eight AI labs, including Mistral AI and Reflection AI. The alliance shares learning data, evaluation, and domain expertise, with the first open model jointly developed by Mistral AI and Nvidia serving as the basis for the Nemotron 4 product line. This collaboration allows Nvidia to leverage the expertise of leading AI labs and accelerate the development of its open models.
Key facts
- Nemotron 4 will have at least 1 trillion parameters, nearly twice the size of Nemotron 3 Ultra (550 billion parameters).
- Nemotron 3.5 Lightning has 30 billion parameters (active 3 billion) and can be operated with a single GPU.
- Nvidia's Nimo Switchyard is an open-source routing tool that allocates requests to the most suitable model, reducing AI token costs.
- The Nemotron Alliance includes eight AI labs, with the first open model jointly developed by Mistral AI and Nvidia.
- Nemotron 3 Ultra scored 48 points on the Intelligence Index, trailing behind top models like Claude Opus 4.8 (61 points) and Kimi K2.6 (54 points).
- Nvidia aims to complete the development of Nemotron 4 as early as late this fall.
- The strategy targets startups and general companies to diversify GPU consumers and drive demand.
Context
Nvidia's focus on open models is part of a broader trend in the AI industry, as companies seek to make AI more accessible and affordable. Open models can be downloaded, modified, and fine-tuned for specific use cases, making them attractive to startups and enterprises looking to leverage AI without the high costs associated with closed models. By driving demand for GPUs and data centers, Nvidia can maintain its leadership in the AI hardware market and continue to innovate in AI chip development.
The development of Nemotron 4 and the expansion of the open AI ecosystem also highlight the growing importance of collaboration in the AI industry. By partnering with leading AI labs through the Nemotron Alliance, Nvidia can leverage the expertise of its partners and accelerate the development of its open models. This collaboration allows Nvidia to stay competitive in the rapidly evolving AI landscape and drive innovation in the open AI ecosystem.